Wyniki wyszukiwania - BazTech

1

Respiratory sound denoising using sparsity-assisted signal smoothing algorithm

Haider Nishi Shahnaj, Behera Ajoy K.

Biocybernetics and Biomedical Engineering

|

2022

|

Vol. 42, no. 2

481--493

EN

Noises are the unavoidable entities which stands as a big barrier in the field of computerized lung sound (LS) based disease diagnosis, as it impairs the quality of LS and therefore greatly misleads the clinical interpretations done based on that. It has numerous sources, of which a few of them could be the noises due to body sounds, environmental noises, power line noises and recording artifacts, which easily contaminates the LS recordings. This paper presents a novel denoising algorithm to eliminate the noises from LS recordings in a more powerful way using Butterworth band-pass filter and sparsity assisted signal smoothing (SASS) algorithm. This study is carried out over LS captured from 80 Chronic Obstructive Pulmonary Disease (COPD), 80 pneumonia and 80 healthy participants in a clinical environment. Each of the recorded LS is denoised using Butterworth band-pass filter and sparse-assisted signal smoothing algorithm. The denoising performance of the proposed algorithm is evaluated on the basis of denoising performance parameters. As per the evaluation of the denoising performance parameters, it is observed that the proposed denoising method suppressed the LS noises with the signal to noise ratio (SNR) of 66.8 dB and with the peak signal to noise ratio (PSNR) of 78.5 dB. The proposed endeavour can be recommended for clinical use for producing noise free LSs to bring effective interpretations. The future endeavours involve the suppression of the heart sound noises from the LS recordings.

2

Computerized lung sound based classification of asthma and chronic obstructive pulmonary disease (COPD)

Haider Nishi Shahnaj, Behera A. K.

Biocybernetics and Biomedical Engineering

|

2022

|

Vol. 42, no. 1

42--59

EN

Diagnostic ambiguity between chronic pulmonary diseases like asthma and Chronic Obstructive Pulmonary Disease (COPD) is very high, as they exhibit similar symptoms, which is the factor responsible for misdiagnosis, leading to heavy deaths every year. To prevent misdiagnosis, some useful work is highly required. This article presents the implementation of a computerized lung sound (LS) based method to classify asthma and COPD cases. The study is conducted on 80 asthma, 80 COPD and 80 healthy LSs. The LS denoising is carried out using empirical mode decomposition (EMD), Hurst analysis and spectral subtraction method. Wavelet entropy (WE) and wavelet packet energy (WPE) features of LS’s are extracted. Various classifiers like support vector machine (SVM), decision tree (DT), k-nearest neighbor (KNN) and discriminant analysis (DA) are accessed to classify healthy, COPD and asthma using WE and WPE features of LS to produce better outcomes. Using the proposed algorithm, the study discriminates between healthy, asthma and COPD cases based on LS with a considerable classification accuracy of 99.3% using the decision tree (DT) classifier. Thus, the study confirms the successful differentiation of asthma and COPD based on LS. Future endeavours will be based on the validation of this algorithm to distinguish the real-time LS data acquired from asthmatic and COPD patients.

3

Comparison of Moving Average and Differential Operation for Wheeze Detection in Spectrograms

Hsueh Meng-Lun, Chen Jin-Peng, Lu Bing-Yuh, Wu Huey-Dong, Liu Pei-Yi

Archives of Acoustics

|

2022

|

Vol. 47, No. 3

383--388

EN

A moving average (MA) is a commonly used noise reduction method in signal processing. Several studies on wheeze auscultation have used MA analysis for preprocessing. The present study compared the performance of MA analysis with that of differential operation (DO) by observing the produced spectrograms. These signal preprocessing methods are not only applicable to wheeze signals but also to signals produced by systems such as machines, cars, and flows. Accordingly, this comparison is relevant in various fields. The results revealed that DO increased the signal power intensity of episodes in the spectrograms by more than 10 dB in terms of the signal-to-noise ratio (SNR). A mathematical analysis of relevant equations demonstrated that DO could identify high-frequency episodes in an input signal. Compared with a two-dimensional Laplacian operation, the DO method is easier to implement and could be used in other studies on acoustic signal processing. DO achieved high performance not only in denoising but also in enhancing wheeze signal features. The spectrograms revealed episodes at the fourth or even fifth harmonics; thus, DO can identify high-frequency episodes. In conclusion, MA reduces noise and DO enhances episodes in the high-frequency range; combining these methods enables efficient signal preprocessing for spectrograms.

4

Transmission Perspective on the Mechanism of Coarse and Fine Crackle Sounds

Lu Bing-Yuh, Hsueh Meng-Lun, Wu Huey-Dong

Archives of Acoustics

|

2021

|

Vol. 46, No. 2

289--300

EN

The possibility of a normal distribution indicates that few particles are in the same phase during a breath and their reflections can be observed on the chest wall, then a few explosive waves with relatively large power occurr occasionally. Therefore, the one-cycle sine wave which is simulated as a single burst of the explosive effect phenomenon penetrates through the chest wall and was analysed to explore the reason of the crackle sounds. The results explain the differences between the definitions of crackle proposed by Sovijärvi et al. (2000a). The crackles in the lungs were synthesised by a computer simulation. When the coarse crackles occur, the results indicate that higher burst frequency carriers (greater than 100 Hz) directly penetrate the bandpass filter to simulate the chest wall. The simulated coarse crackle sounds were low pitched, with a high amplitude and long duration. The total duration was greater than 10 ms. However, for a lower frequency carrier (approximately 50 Hz), the fundamental frequency component was filtered out. Therefore, the second harmonic component of the lower frequency carrier, i.e., the fine crackle, penetrated the chest wall. Consequently, it is very possible that the normal lung sounds may contain many crackle-shaped waves with very small amplitudes because of the filtering effects of the chest wall, environment noises, electric devices, stethoscopes, and human ears, the small crackles disappear in the auscultations. In addition, our study pointed out that some unknown crackles of the very low frequency under the bandwidth of the human ears cannot penetrate the airways and be detected by medical doctors. Therefore, it might be necessary to focus advanced electronic instrumentation on them in order to analyse their possible characteristics for diagnosis and treatment of the respiration system.

5

Machine learning in lung sound analysis: a systematic review

Palaniappan R., Sundaraj K., Ahamed N. U.

Biocybernetics and Biomedical Engineering

|

2013

|

Vol. 33, no. 3

129--135

EN

Machine learning has proven to be an effective technique in recent years and machine learning algorithms have been successfully used in a large number of applications. The development of computerized lung sound analysis has attracted many researchers in recent years, which has led to the implementation of machine learning algorithms for the diagnosis of lung sound. This paper highlights the importance of machine learning in computer-based lung sound analysis. Articles on computer-based lung sound analysis using machine learning techniques were identified through searches of electronic resources, such as the IEEE, Springer, Elsevier, PubMed and ACM digital library databases. A brief description of the types of lung sounds and their characteristics is provided. In this review, we examined specific lung sounds/disorders, the number of subjects, the signal processing and classification methods and the outcome of the analyses of lung sounds using machine learning methods that have been performed by previous researchers. A brief description on the previous works is thus included. In conclusion, the review provides recommendations for further improvements.